Am Mo 31. März 2008 schrieb Andy Green: [...] > This patch protects the acknowldge / poll ISR with a spinlock that defers > interrupt acknowledge by the processor until the service for the first > accelerometer is completed. >
Isn't this common best practice for ISRs to lock the own interrupt (and possibly lower prioritized interrupts of other devices (AccMeter2)) at very beginning of ISR, prior to resetting it and handling event. At very end of ISR arm it (them) again. Z80 even had a RETI command for this. This is not only to cope with level triggered IRQs, but also to avoid infinite reentrant nesting when CPU ISR-turnaround is slower than IRQ-freqency, even with edge-triggered IRQs. So no need for a spinlock then? cheers jOERG
